When Candidates Attack: Who Goes Negative, and Why it Works

Citation

Mattes, Kyle Alan (2008) When Candidates Attack: Who Goes Negative, and Why it Works. Dissertation (Ph.D.), California Institute of Technology. doi:10.7907/RPKP-J818. https://resolver.caltech.edu/CaltechETD:etd-05302008-130905

Abstract

Political candidates use a variety of negative campaigning strategies, and these attacks have different degrees of success. To explain this, I first introduce a formal model of campaign strategy to show when candidates will engage in negative campaigning and how it can affect election results. Whether candidates choose negative campaigning depends upon three factors: the voters' preconceptions about political candidates' the voters' preferred dimension, and the candidates' character traits. I show that eliminating negative campaigning has an ambiguous effect on voter welfare. Then, I extend the formal model of campaigning to include situations in which candidates campaign dishonestly. I find that allowing this is strictly inferior for voter welfare. Often this is because whenever lying becomes more probable, all campaigns start to look the same.

In order to understand the implications of the formal model, I use multiple empirical testing methods. I use laboratory experiments to test how voter beliefs and voter behavior affect the frequency and content of candidates' negative campaigning. I find support for the above comparative statics hypothesis about when candidates are more likely to negatively campaign. Furthermore, the laboratory results are better explained by a quantal response equilibrium model in which voters are risk averse and are more naive than sophisticated in their Bayesian updating ability.

I next consider how negative campaigning translates into votes. From survey responses, I show empirical support for candidate tactics that target specific segments of voters based upon their predilections toward accepting negativity in campaigns. One way candidates attack their rivals is by trying to make voters afraid of them. From a laboratory experiment in which subjects made social judgments about pairs of candidates who ran against each other in real elections, we found that the candidates chosen by subjects as more physically threatening had actually lost 65% of the real elections. Such correlation could arise from common neural mechanisms engaged when viewers make social judgments about faces, and when voters evaluate candidates. This was shown in two independent fMRI studies in which subjects again made judgments about images of unfamiliar politicians who had run in real elections. We found that for judgments of threat, election loss correlated with activity in the insula and ventral anterior cingulate cortex, structures known to be involved in processing negative emotions. These findings suggest that voters are influenced by negative emotions elicited by a candidate’s mere appearance.

Finally, we examine the role of candidate fear in explaining voter choice in presidential elections using ANES data from 1980-2004. We find that certain groups of voters are more susceptible to fear appeals. After controlling for the factors commonly thought to affect voter decision-making, we also find further evidence that fear of presidential candidates directly affects vote choice. The results have implications for future research on negative political communication, as they suggest that threatening first impressions can harm a candidate's electoral chances. They also show that candidates who wish to successfully utilize negative campaigning must tailor their campaigns not only to reflect the weaknesses of the opponent but also to insure that their message is salient to the specific group of targeted voters.

Thesis Files